MethylPipeR: MethylPipeR

library(bigmemory)

tmpSessionLogFolder <- paste0(tempdir(), "/")

initLogs(sessionLogFolder = tmpSessionLogFolder, note = "Test note.")

dummyTrainXs <- matrix(rnorm(1000), ncol = 10)
dummyTrainY <- rbinom(100, 1, 0.2)

dummyTestXs <- matrix(rnorm(1000), ncol = 10)
dummyTestY <- rbinom(100, 1, 0.2)

dummyCovariatesTable <- matrix(rnorm(300), ncol = 3)

# Test binary models
binaryglmnetModel <- fitMPRModel(type = "binary", method = "glmnet", trainXs = dummyTrainXs, trainY = dummyTrainY, alpha = 0.5)
expect_s3_class(binaryglmnetModel, "MPRModel")
expect_s3_class(binaryglmnetModel$model, "glmnet")

binarybiglassoModel <- fitMPRModel(type = "binary", method = "biglasso", trainXs = dummyTrainXs, trainY = dummyTrainY, alpha = 0.5)
expect_s3_class(binarybiglassoModel, "MPRModel")
expect_s3_class(binarybiglassoModel$model, "biglasso")

binarybartModel <- fitMPRModel(type = "binary", method = "bart", trainXs = dummyTrainXs, trainY = dummyTrainY)
expect_s3_class(binarybartModel, "MPRModel")
expect_s3_class(binarybartModel$model, "pbart")

binaryrfModel <- fitMPRModel(type = "binary", method = "rf", trainXs = dummyTrainXs, trainY = dummyTrainY)
expect_s3_class(binaryrfModel, "MPRModel")
expect_s3_class(binaryrfModel$model, "randomForest")

# Test continuous models
continuousglmnetModel <- fitMPRModel(type = "continuous", method = "glmnet", trainXs = dummyTrainXs, trainY = rnorm(100), alpha = 0.5)
expect_s3_class(continuousglmnetModel, "MPRModel")
expect_s3_class(continuousglmnetModel$model, "glmnet")

continuousbiglassoModel <- fitMPRModel(type = "continuous", method = "biglasso", trainXs = dummyTrainXs, trainY = rnorm(100), alpha = 0.5)
expect_s3_class(continuousbiglassoModel, "MPRModel")
expect_s3_class(continuousbiglassoModel$model, "biglasso")

continuousbartModel <- fitMPRModel(type = "continuous", method = "bart", trainXs = dummyTrainXs, trainY = rnorm(100))
expect_s3_class(continuousbartModel, "MPRModel")
expect_s3_class(continuousbartModel$model, "wbart")

continuousrfModel <- fitMPRModel(type = "continuous", method = "rf", trainXs = dummyTrainXs, trainY = rnorm(100))
expect_s3_class(continuousrfModel, "MPRModel")
expect_s3_class(continuousrfModel$model, "randomForest")

# Test survival models
dummySurvTrainY <- data.frame(list(
  "time_to_event" = runif(100, 0, 10),
  "Event" = dummyTrainY
))
dummySurvTestY <- data.frame(list(
  "time_to_event" = runif(100, 0, 10),
  "Event" = dummyTestY
))

survivalglmnetModel <- fitMPRModel(
  type = "survival", method = "glmnet",
  trainXs = dummyTrainXs, trainY = dummySurvTrainY,
  tteColname = "time_to_event", eventColname = "Event"
)
survivalbiglassoModel <- fitMPRModel(
  type = "survival", method = "biglasso",
  trainXs = dummyTrainXs, trainY = dummySurvTrainY,
  tteColname = "time_to_event", eventColname = "Event"
)
survivalrfModel <- fitMPRModel(
  type = "survival", method = "rf",
  trainXs = dummyTrainXs, trainY = dummySurvTrainY,
  tteColname = "time_to_event", eventColname = "Event"
)

expect_s3_class(survivalglmnetModel, "MPRModel")
expect_s3_class(survivalglmnetModel$model, "glmnet")

expect_s3_class(survivalrfModel, "MPRModel")


# Test survival columns not matching those specified
expect_error(do.call(fitMPRModel,
  type = "survival", method = "glmnet",
  trainXs = dummyTrainXs, trainY = dummySurvTrainY,
  tteColname = "tte", eventColname = "Event"
))
expect_error(do.call(fitMPRModel,
  type = "survival", method = "glmnet",
  trainXs = dummyTrainXs, trainY = dummySurvTrainY,
  eventColname = "event"
))
colnames(dummySurvTestY) <- c("tte", "Event")
expect_error(do.call(fitMPRModel,
  type = "survival", method = "glmnet",
  trainXs = dummyTrainXs, trainY = dummySurvTrainY,
  testXs = dummyTestXs, testY = dummySurvTestY
))
colnames(dummySurvTestY) <- c("time_to_event", "event")
expect_error(do.call(fitMPRModel,
  type = "survival", method = "glmnet",
  trainXs = dummyTrainXs, trainY = dummySurvTrainY,
  testXs = dummyTestXs, testY = dummySurvTestY
))

# Test NA checking
dummyTestY[[10]] <- NA
expect_error(do.call(fitMPRModel,
  type = "binary", method = "glmnet",
  trainXs = dummyTrainXs, trainY = dummyTrainY,
  testXs = dummyTestXs, testY = dummyTestY, alpha = 0.5
))
dummyTrainXs[10, 10] <- NA
expect_error(do.call(fitMPRModel,
  type = "binary", method = "glmnet",
  trainXs = dummyTrainXs, trainY = dummyTrainY,
  testXs = dummyTestXs, testY = dummyTestY, alpha = 0.5
))

dummyTrainY[[10]] <- NA
expect_error(do.call(fitMPRModel,
  type = "binary", method = "glmnet",
  trainXs = dummyTrainXs, trainY = dummyTrainY,
  testXs = dummyTestXs, testY = dummyTestY, alpha = 0.5
))

dummyTrainXs[10, 10] <- NA
expect_error(do.call(fitMPRModel,
  type = "binary", method = "glmnet",
  trainXs = dummyTrainXs, trainY = dummyTrainY,
  testXs = dummyTestXs, testY = dummyTestY, alpha = 0.5
))